Author: Sandro Necchi
Date: 13:10:54 02/20/06
Go up one level in this thread
On February 20, 2006 at 12:14:01, Vincent Diepeveen wrote: >On February 19, 2006 at 05:00:48, Sandro Necchi wrote: > >I feel you 'hit the nail on the head' when predicting that in 2006 we will see >huge progresses in engines. I did not say that I am the only one to think this, but some people do not and feel that what is true today will be the same tomorrow... The message was toward them... >A result is that certain engines jump harder than >others. So real strong engines of the past will fade and new engines will jump >bigtime. Well, I believe some engine of the past will do, but some will keep up and be able to improve further...I have no data on new ones, so I am expecting some one to come out from the dark, like Rybka, but I am not sure which one... > >Fruit's public source code has already proven that. It has boosted in a direct >manner especially engines like Spike, Toga, not to mention Rybka. Yet it's >public code so we can see where the jump came from and learn from that for our >own engines. Yes, this is known...but I was referring also to the possibility to face a stronger opponent to get more data where to make improvements...if one is able to understand this, of course... > >What is not sure is which engines make the biggest jump in terms of elo >strength. In openingsbooks things are still clear, simply because there is only >1 professional near to fulltime openingsbook creator on this planet (Erdogan >Gunes) who in world champs 2006 gets the support of a 1056 processor Altix (from >which zappa effectively will search on a cpu or 450-500 i guess as largest >partition is 512p). Yet he this time doesn't get help of a titled player for his >openings book. In my predictions I was referring to the SSDF list and not the 2006 WCCC. Of course with such an hardware and bugfree that program has the biggest chances, but on a few games the luck factor (correct openings choice against specific opponents...good positions out of book even if taken out of book soon...) can play a lot to the final score... > >On the other hand there is Noomen now who can make a book for Rybka in world >champs 2006 Turino, which will for sure be a book full of fireballs. If you >stumble into one of such fireballs then you're completely dead. I know it as I know him since nearly 20 years now... > >I personally doubt however Fritz can make a big jump in 2006 based upon only >engine strength. The strong point from Fruit was basically already a strong >point of Fritz and as Frans isn't joining any tournaments anymore with Fritz in >future, he'll probably have problems finding information how to improve his >program, whereas at world champs 2005 there were very interesting talks in this >respect where programmers could learn from other programmers. I was referring to the SSDF list and I think another gain is possible, but not more than 50-70...of course it could be less as well...but this would place the program away from the first ones... > >Online commercial programmers simply aren't talking. > >In short other programmers such as me are learning a lot more now than Frans >can, despite his amazing understanding of how IDA pro works (to debug Quest of >course). Yes, I know you have a lot of potential, but since I was referring to the SSDF list and I have not news of your program being commercial I did not mention you. > >It's true that on paper Fruit could progress really a lot, yet Fabien will face >a few practical problems which were similar to when i was making a go playing >program. I stopped its development for the same reason and other French >programmers also have that same problem. So Fabien still has to prove itself >there. Yes, it is not clear, but I personally think he will be able to improve further. > >Making something simplistic that's real well tuned and is kicking unprecedented >butt is a true historic achievement. Yet Fruit progress could only come from >points now where Tiger and Hiarcs also have problems progressing. So this is an >open challenge where end of 2006 we will be able to give an answer too. World >champs 2006 will really come too soon in that respect. Well, the 2006 WCCC will be fun and to me is not clear who will be the winner...we will see... I believe Hiarcs will improve further as the programmer is quite good and now working full time on it... > >I also look forward to the new DeepSjeng. I remember how in a previous lifetime >its unique playstyle was real interesting. It seems that multiprocessing is >completely fixed now and superior working in comparision to older versions. He has to make a big jump to me to be really competitive for the title... > >It is here where engines like Rybka, Fruit, still have to prove itself. Well, yes the SMP version of Rybka would be quite strong...especially with a strong book... > >Yet you really are forgetting to mention a bunch of amateur engines. Starting >with Glaurung, The Baron, Spike and so on. Well, I did not forget...I have no data for their commercial version (SSDF list) > >Glaurung is nearly SMP now. Just in time for world champs. The Baron already was >and latest revisions of The Baron where Diep of course tests a lot against, seem >really hundreds of rating points improved. OK, but I had not data on this one... > >From shredder10 i am sure it will be improved a lot. He has proven that in the >past and he'll do that in the future once again. Yes, I can see it... > >The real interesting thing of 2006 is which new engines will be there besides >Rybka that can really deliver a hard punch to opponents. I think the level got closer, so it will be a hard figth... > >Just improving 100 points simply isn't enough for that. I agree. > >>On February 18, 2006 at 17:20:59, stuart taylor wrote: >> >>>On February 17, 2006 at 12:55:10, Sandro Necchi wrote: >>> >>>>On February 17, 2006 at 12:03:28, Marc Lacrosse wrote: >>>> >>>>>On February 17, 2006 at 11:44:05, Sandro Necchi wrote: >>>>> >>>>>>On February 17, 2006 at 04:55:08, John Jack wrote: >>>>>> >>>>>>>Shredder 10 is due two be released in April. Will it be stronger than Rybka??? >>>>>>> >>>>>>>J.J >>>>>> >>>>>>Shredder 10 is planned to be released in March and not April. >>>>>> >>>>>>For the second question it is not easy to answer as Rybka commercial version is >>>>>>not available yet and Shredder 10 is not ready yet (under development still), so >>>>>>you will have to wait March to find out. >>>>>> >>>>>>Sandro >>>>> >>>>>Thanks for the information Sandro ! >>>> >>>>you welcome. >>>> >>>>> >>>>>By the way, will there still be two different versions (UCI and CB) ? >>>> >>>>Nothing has been decited yet as nothing has been shown to CB yet (that I know >>>>of) >>>> >>>>We are still testing new better versions all the time. >>>> >>>>However my personal opinion is yes because CB will make it possible...I let you >>>>guess why... >>> >>>So it's coming on well! Well, so is Rybka, I guess. It's going going to be >>>tough. All top programmers have to again become like very young in energy and >>>determination. >>>OK, so Shredder MIGHT do it, Rybka we know. Hiarcs and Fritz are what they are, >>>Junior, silent, Tiger, off the scene, Fruit, looks like it did what it could, >>>Chess master, never were too ambitious, others retired, I don't know? >> >>First of all I am not a magician, so the following expectations are based on my >>knowledge of the programs and the way they are made as well as opinion on the >>programmers ability and williness to work on their engines... >> >>Yes, Shredder has been improved a lot, but we are still working on it, so the >>final version is not available yet and one cannot tell how strong it will be... >> >>Rybka we all know...expectations are easy for this program. >> >>Hiarcs is quite good and surely has more room for improvements, but his >>programmer, which is one of the best, is working part time so we cannot expect >>another big jump in the near future. I could be wrong of course if he can spend >>a lot of time improving the engine as he has the potential... >> >>Fritz can be improved further and I am expecting another 50-70 Elo gain. Why so >>much if normally it did not improved that much? Because (and this is for all >>programs) the info you can get from a much stronger program let you speed up the >>improvements being able to see more clearly the weaknesses of your program. >> >>Junior has a lot of potential too, but it seems the programmer is more busy with >>other stuff...it can improve a lot if he can/wants to, but I doubt it would be >>more than 100 Elo points... >> >>Tiger has potential too, but it looks like his programmer is taking a break, so >>if this is true will not improve in the near future. I would expect a gain of >>50-70 Elo if he does work on the engine...he shall work on the SMP version as >>well... >> >>Fruit has more potential as it is still a young program and to expect 80-100 Elo >>improvements is not being too optimistic...his programmer is quite good. >> >>Zappa will improve further as the version we saw was weaker than the previous >>one. It is already a very strong program...maybe we will see a commercial >>version too, but not in the near future it seems... >> >>Chess Master is already strong and can be improved further, but it seems that >>his programmer is not interested on working on the engine as well as supply it >>with a professional opening book...the programmer is quite good and may change >>his mind...nobody seems to know... >> >>Gandalf is strong and seems to have a lot of room for improvements...his >>programmer is quite experienced and seems motivated to get to the top...this is >>one of the program I would expect the highest improvements... >> >>There are other programs which are quite strong and that can improve further...I >>would not forget some programmers that are not active now, but were before and >>that may will come back... >> >>2006 looks quite fun for computer chess funs! >> >> >>Sandro >>> >>>S.Taylor >>>> >>>>> >>>>>Marc Sandro
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.